Unsupervised Personal Name Disambiguation

Unsupervised Personal Name Disambiguation Gideon S. Mann and David Yarowsky Department of Computer Science Johns Hopkins University Baltimore, MD 21218 USA {gsm,yarowsky}@cs.jhu.edu Abstract This paper presents a set of algorithms for distinguishing personal names with mul- tiple real referents in text, based on little or no supervision. The approach utilizes an unsupervised clustering technique over a rich feature space of biographic facts, which are automatically extracted via a language-independent bootstrapping pro- cess. The induced clustering of named entities are then partitioned and linked to their real referents via the automatically extracted biographic data. Performance is evaluated based on both a test set of hand- labeled multi-referent personal names and via automatically generated pseudonames. 1 Introduction One open problem in natural language ambiguity resolution is the task of proper noun disambigua- tion1. While word senses and translation ambigu- ities may typically have 2-20 alternative meanings that must be resolved through context, a personal name such as “Jim Clark” may potentially refer to hundreds or thousands of distinct individuals. Each different referent typically has some distinct contex- tual characteristics. These characteristics can help distinguish, resolve and trace the referents when the surface names appear in online documents. A search of Google shows 76,000 web pages mentioning Jim Clark, of which the first 10 unique referents are: 1This has been recognized even by the popular press. Reuters (March 13, 2003) observed the problem of name am- biguity to be a major stumbling block in personal name web searches. 1. Jim Clark - Race car driver from Scotland 2. Jim Clark - Clockmaker from Colorado 3. Jim Clark - Film Editor 4. Jim Clark - Netscape Founder 5. Jim Clark - Disaster Survivor 6. Jim Clark - Car Salesman in Kansas 7. Jim Clark - Fishing Instructor in Canada 8. Jim Clark - Computer Science student in Hong Kong 9. Jim Clark - Professor at McGill 10. Jim Clark - Gun Dealer in Louisiana In this paper, we present a method for dis- tinguishing the real world referent of a given name in context. Approaches to this problem include Wacholder et al. (1997), focusing on the variation of surface name for a given referent, and Smith and Crane (2002), resolving geographic name ambiguity. We present preliminary evaluation on pseudonames: conflations of multiple personal names, constructed in the same way pseudowords are used for word sense disambiguation (Gale et al., 1992). We then present corroborating evidence from real personal name polysemy to show that this tech- nique works in practice. Miles Davis birth day May 26(5), May 25(5) birth year 1926(82), 1967(18), 1969(9)... occupation trumpeter(38), artist(10), player(5)... birth place Alton(7), Illinois(6) Joerg Haider birth year 1950(6) occupation leader(198) politician(93) chairman(6)... birth place Austria(1) Table 1: Extracted Biographical Information from 1000 Web Pages Another topic of recent interest is in producing biographical summaries from corpora (Schiffman et al., 2001). Along with disambiguation, our system simultaneously collects biographic information (Ta- ble 1). The relevant biographical attributes are de- picted along with a clustering which shows the dis- …

Content truncated to 500 words